top of page

Method of Working

BCS approach.jpeg

​1. Defining the problem

We start with the big question - improving customer service. Within the big question together with the client we define the areas of improvement in the customer journey relevant to their hospitality establishment. We always take into the consideration clients’ values and brand promise.

 

2. Collection of data

The next step is identifying  available data. AI needs a lot of data to work. This could be anything from property descriptions, manuals, summaries, itineraries, customer reviews, prices, timings, etc. The key is to gather data that's useful and relevant to the problem and areas set for improvement.

 

We add to the mix the technical and regulatory restrictions that define the overall framework for how data can be collected and analysed.

For example: 'For the VCSA Tester included into our presentation for Omena Hotels (Finland), we have gathered data available on their website, their application, and their expedia entries. This included descriptions of rooms and amenities, booking process, history of the Omena Hotels, and links to partners' site.'

 

3. Cleaning and Preparing the Data

Data isn’t always neat. Before it can be used it needs to be cleaned up: errors to be removed, missing information to be filled in, and making sure it’s in the right format for AI to understand.

​For example: 'While preparing data for the VCSA Tester included into our presentation for SOKOS Hotels (Finland), we reorganised, structured, and edited information gathered from SOKOS Hotels website, creating a separate file that AI Assistant can understand and work with.'

 

4. Building the Model

At this stage the right algorithm is chosen and the AI model is built that will learn from the client’s data. This is where AI learns patterns and starts to make predictions or decisions.

For example: 'Both for Omena and SOKOS Hotels we tested different LLMs, such as Gemini 1.5 Pro, Claude 3 Haiku, Claude 3.5 Sonnet, and GPT 4.0 and chose Claude Sonnet.'

 

5. Testing of the Model

Once the model is built, it needs to be tested. We’ll see how well it performs on new data. If it doesn’t do well, adjustments are made, and the model is fine-tuned.

For example: 'The VCSA Tester included into our presentation for SOKOS Hotels (Finland) we tested by providing for the chosen LLM, Claude 3.5 Sonnet, the data related to SOKOS Hotels we have collected and prepared. The testing consisted of asking various questions in a number of languages, and checking the answers against the data input.'

 

6. Deploying the Model

When the model is ready, it’s time to put it to work in the real world. Whether it’s in a customer-facing app, an internal tool, or an automated process, this is where AI starts delivering value.

 

7. Monitoring and Improving

AI isn’t just set and forget. We’ll need to keep an eye on how it performs, update it with new data, and keep improving it over time.

bottom of page